Design changes in large-scale engineering projects often propagate across interdependent components, posing significant risks to project cost, schedule, and quality. Understanding how a single change may trigger downstream modifications is critical for effective change management. This study proposes a novel data-driven methodology for modeling the causal structure of change propagation. By comparing geometric data from the initial and final design models and applying the Fast Causal Inference (FCI) algorithm, we construct a statistically grounded change propagation network that captures the underlying dependencies among structural elements. Conditional probabilities are estimated to quantify the likelihood of change propagation across components. Unlike conventional approaches that rely on expert-defined pairwise dependencies or static design rules, the proposed method offers an objective, scalable framework that leverages readily available design documentation. We demonstrate the feasibility of the approach using data from a real-world infrastructure project, revealing both intra- and inter-category causal relationships. This provides a foundation for developing adaptive project management tools that enable automated risk detection and proactive change control across a wide range of engineering domains.